🤖 Self-Evolving AI Agent

🌟 A self-evolving AI agent that analyzes tasks, searches tools, generates code, and executes solutions autonomously.

🚀 What's New in V2

• ⚡ 1000x Faster Search: Keyword-based search replaces embeddings (5-10s → <5ms)
• 🌏 Chinese Support: Full support for Chinese queries and responses
• 📦 Smart Dependencies: Prompts before installing packages with clear feedback
• 🔍 Better Tool Discovery: 100% search accuracy with enhanced keyword matching

✨ Key Features

• 🧠 Smart Task Analysis - Breaks complex tasks into actionable steps
• 🔍 GitHub Tool Search - Finds and integrates relevant tools automatically
• 🛠️ Code Generation - Creates reusable, generic tools
• 💾 Tool Memory - Learns and saves tools for future use
• 🌏 Bilingual - Works seamlessly with English and Chinese

🛠️ Quick Start

1. Clone & Install

git clone https://github.com/Ramond-e/Self-evolutional-agent
cd Self-evolutional-agent
pip install -r requirements.txt

2. Configure API Keys

Create a .env file with at least ONE LLM key. You can add both (OpenRouter优先):

OPENROUTER_API_KEY=your_openrouter_key_here        # (优先使用)
OPENROUTER_API_BASE_URL=https://openrouter.ai/api/v1/chat/completions
OPENAI_API_KEY=your_openai_key_here                # (如未填OpenRouter则用此)
OPENAI_API_BASE_URL=https://api.openai.com/v1/chat/completions
GITHUB_API_TOKEN=your_github_token_here

API KEY与BASE URL优先级与变量说明：

• OPENROUTER_API_KEY 和 OPENROUTER_API_BASE_URL 均需设置才能使用 OpenRouter。
• OPENAI_API_KEY 和 OPENAI_API_BASE_URL 均需设置才能使用 OpenAI（API BASE URL默认即可，需特殊代理等可自定义）。
• 两类 key & base url 都填写时，优先调用 OpenRouter。
• 只填写其中一组就用填写的。
• 都没填会报错。

3. Run

python manager.py

🎮 Usage Examples

Stock Price Query

💬 User: 查看英伟达的股票价格

📋 Alita breaks it down:
  1. [✓] Fetch NVIDIA stock data
  2. [✓] Analyze price performance  
  3. [✓] Present results in Chinese

📊 Result: 英伟达(NVDA)当前股价为$131.26，较昨日下跌1.17%...

Weather Query

💬 User: 北京今天的天气怎么样？

🌤️ Result: 北京今天天气晴朗，温度28°C，湿度70%...

Special Commands

• list tools/ - Show all available tools
• exit - Quit the application

🏗️ Architecture

alita/
├── manager.py          # Main controller
├── task_analyzer.py    # Task decomposition
├── tools_managing.py   # Tool storage (JSON + keywords)
├── code_generating.py  # Code generation
├── code_executing.py   # Execution engine
├── tool_searching.py   # GitHub integration
└── tools/             # Saved tools (JSON format)

📊 Technical Improvements

Tool Storage Format

{
  "id": "weather_tool_1234567890",
  "tool_description": "Fetches weather data for any location",
  "keywords": "weather temperature forecast climate 天气 温度",
  "install_dependencies": ["pip install requests"],
  "python_code": "..."
}

Search Algorithm

• Keyword Matching: Direct keyword intersection (40%)
• Description Matching: Keywords in description (20%)
• Partial Matching: Substring matches (20%)
• Domain Boost: Weather/Stock specific bonuses

🤝 Contributing

1. Fork the repository
2. Create your feature branch
3. Commit your changes
4. Push to the branch
5. Open a Pull Request

📝 License

MIT License - see LICENSE file for details.

🙏 Acknowledgments

• OpenRouter for AI model access
• GitHub for code repository hosting
• Open source community for inspiration

---

⭐ Star this repo if you find it useful!